Correct Answer: B. Unpaired T test
The unpaired (independent) t-test is the gold standard for comparing means between two independent groups where one is a control. This test is appropriate when: (1) the two groups are independent/unrelated (e.g., treatment vs. control arm in a randomized trial), (2) the outcome is continuous/quantitative (e.g., blood pressure, hemoglobin, BMI), and (3) sample sizes are small to moderate (typically n < 30 per group, or when population SD is unknown). The unpaired t-test assumes normality of the outcome variable and equal variances between groups (Levene's test confirms this). It calculates the t-statistic as the difference in group means divided by the pooled standard error, then compares against the t-distribution with (n₁ + n₂ − 2) degrees of freedom. In Indian clinical trials (e.g., comparing a new antihypertensive agent against placebo in a randomized controlled trial), this is the routine statistical test reported. The unpaired t-test is more powerful than non-parametric alternatives (Mann-Whitney U) when normality assumptions hold, making it the first choice for hypothesis testing in comparative studies with a control group.
Why the other options are wrong
A. Chi-square test — Chi-square is used for categorical/nominal data (e.g., disease present/absent, treatment success/failure), not continuous outcomes. It tests association between two categorical variables, not difference in means. This is a common trap—students confuse 'comparing groups' with 'comparing means.' Chi-square would be appropriate only if the outcome were categorical (e.g., cure vs. no cure). C. Paired T test — Paired t-test is used when the same subjects are measured twice (before-after, matched pairs, crossover design), not for independent groups. A control group is inherently independent of the treatment group. Paired t-test assumes dependency/correlation between observations, which violates the independence assumption here. This trap confuses 'two groups' with 'two measurements.' D. Z test — Z-test is appropriate only when the population standard deviation is known and sample size is large (n > 30). In most clinical studies, population SD is unknown, so we use the sample SD and t-distribution instead. Z-test is rarely used in practice; t-test is more conservative and appropriate for small-to-moderate samples typical in Indian clinical research.
High-Yield Facts
- Unpaired t-test: compares means of two independent groups; assumes normality, equal variances, and continuous outcome.
- Chi-square test: for categorical data; tests association, not difference in means.
- Paired t-test: for dependent/matched observations (same subjects measured twice); NOT for independent control vs. treatment groups.
- Z-test vs. t-test: Z-test requires known population SD and large n (>30); t-test uses sample SD and is standard for small-to-moderate samples.
- Degrees of freedom for unpaired t-test: df = n₁ + n₂ − 2; determines critical t-value from t-distribution table.
Mnemonics
INDEPENDENT = UNPAIRED T If two groups are INDEPENDENT (control vs. treatment, no matching), use UNPAIRED t-test. If same subjects measured twice (before-after), use PAIRED t-test. If outcome is categorical (yes/no), use CHI-SQUARE. T-TEST vs Z-TEST: Know Your SD T-test: population SD unknown (use sample SD), small-moderate n. Z-test: population SD known, large n (>30). In clinical practice, you almost never know population SD → use t-test.
NBE Trap
NBE pairs "two groups" with "chi-square" to trap students who confuse 'comparing groups' with 'comparing categorical outcomes.' The key discriminator is the type of outcome variable (continuous vs. categorical), not merely the presence of two groups.
Clinical Pearl
In a typical Indian RCT (e.g., comparing a new tuberculosis regimen vs. standard RNTCP regimen on sputum conversion time in weeks), you measure a continuous outcome (time in days) in two independent groups. The unpaired t-test directly answers: "Is the mean conversion time significantly different between the two arms?" This is the routine statistical test reported in Indian medical journals and regulatory submissions.
_Reference: Park's Textbook of Preventive and Social Medicine, Ch. 10 (Biostatistics); Harrison's Principles of Internal Medicine, Ch. 5 (Statistical Interpretation of Data)_